Th2 bias and T-cell exhaustion characterize the immunopathology of non-tuberculous mycobacterial pulmonary disease

Abstract

Non-tuberculous mycobacterial pulmonary disease (NTM-PD) is an escalating global health concern with poorly defined immunological mechanisms, necessitating comprehensive profiling to guide therapeutic advances. We analyzed peripheral blood from 28 treatment-naïve NTM-PD patients (19 Mycobacterium avium complex, 9 Mycobacterium abscessus) and 27 matched controls using 42-marker mass cytometry (CyTOF) and Luminex multiplex assays. A random forest model identified predictive markers, while an in vitro murine macrophage model evaluated chemokine production. NTM-PD patients displayed significant immune shifts, including increased classical monocytes (CD14+ CD16−), reduced NKT-like cells (CD3+ CD56+), and elevated T-cell exhaustion markers (PD-1, TOX). This coincided with a Th1/Th2 balance shift characterized by heightened IL-13. Elevated IFN-γ-inducible chemokines CXCL9 and CXCL10 coexisted with this Th2-biased signature, indicating a complex, dysregulated inflammatory state. A model integrating immune-cell frequencies and cytokine profiles achieved robust diagnostic accuracy (AUC = 0.922) with prognostic potential. In vitro, NTM-infected macrophages produced substantial CXCL9 and CXCL10 levels relative to the LPS maximal activation benchmark, identifying them as a major cellular source. These findings propose an immunological framework wherein T-cell exhaustion and a Th2-biased microenvironment strongly correlate with NTM-PD pathogenesis. CXCL9, CXCL10, and IL-13 emerge as candidate therapeutic targets, while our predictive model offers a foundational approach for risk stratification.

Introduction

Nontuberculous mycobacterial pulmonary disease (NTM-PD) has emerged as a significant public health challenge due to its rising global incidence, chronicity, treatment difficulty, and substantial morbidity^1–6. Compared to Mycobacterium tuberculosis infection, NTM-PD exhibits distinct clinical and radiographic features, characterized by bronchiectasis, progressive pulmonary lesions, and variable responsiveness to antimycobacterial therapy^1,7–9. Historically, a subset of NTM-PD cases, known as “Lady Windermere syndrome”, predominantly affects slender, elderly women without clear immunodeficiency¹, but the immune abnormalities driving susceptibility are unclear. Only a fraction of those exposed develop disease, implying that immune dysregulation underlies progression^10,11.

A Th1-type immune response (IFN-γ–mediated macrophage activation) is critical for mycobacterial control¹. By contrast, Th2 immunity (e.g., IL-4, IL-5, IL-13) promotes humoral responses and tissue remodeling but is ineffective against intracellular mycobacteria¹². Even a partial shift from Th1 to Th2 severely impairs mycobacterial control^13,14. Notably, IL-13–driven alternative macrophage (M2) polarization is linked to fibrosis and poor bacterial clearance¹⁵. However, whether such Th1/Th2 imbalance occurs prominently in NTM-PD remains unclear.

Apart from Th1/Th2 skewing, recent studies underscore the critical role of immune exhaustion in chronic infectious diseases. Persistent antigenic stimulation, as encountered in chronic viral infections and cancers, induces sustained expression of inhibitory receptors (e.g., PD-1, CTLA-4, and TIM-3) on T cells, resulting in functional impairment and immune evasion by pathogens^16,17. Additionally, TOX (thymocyte selection-associated high mobility group box), a critical transcription factor sustaining the exhausted phenotype, has recently been recognized as an essential driver of exhaustion in chronic infections and cancer by promoting persistent expression of inhibitory receptors and preventing exhausted T cells from regaining effector functions^18–20. In chronic pulmonary infections caused by mycobacteria, exhausted T cells exhibit impaired cytokine production, reduced proliferative capacity, and compromised ability to orchestrate effective antibacterial responses^21,22. Yet, comprehensive evidence describing immune exhaustion signatures and their clinical relevance in NTM-PD is limited¹⁰.

Recent advances in immune profiling technologies, particularly cytometry by time-of-flight (CyTOF), provide unprecedented capacity to delineate complex immune signatures at single-cell resolution^23,24. Studies employing CyTOF have successfully revealed detailed immune alterations in various infectious and inflammatory diseases, identifying novel pathogenic cell subsets and cytokine profiles previously unrecognized using traditional flow cytometry^25,26. Utilizing CyTOF to comprehensively map immune dysfunction in NTM-PD could unveil critical immune markers associated with disease susceptibility and severity, potentially informing more precise therapeutic strategies.

Furthermore, the chemokine axis CXCL9/CXCL10–CXCR3 has recently gained attention as a key mediator in infectious diseases, orchestrating immune-cell recruitment and activation^27,28. Elevated CXCL9/CXCL10 levels are often indicative of ongoing IFN-γ–mediated immune responses in tuberculosis (TB), effectively recruiting Th1 cells for pathogen control^14,29. However, paradoxically elevated CXCL9/CXCL10 in the context of impaired Th1 or enhanced Th2 responses could contribute to a pathogenic cycle of excessive immune-cell influx, activation, and subsequent exhaustion, leading to chronic unresolved inflammation and tissue damage³⁰. Whether a similar scenario exists in NTM-PD, wherein the CXCL9/CXCL10 axis contributes to immunopathology, remains unknown and merits investigation.

In summary, NTM-PD is likely governed by a unique immunopathological framework that differs from TB in key ways. While chronic infections like TB are known to induce T-cell exhaustion, this typically occurs in the context of a strong, albeit ultimately overwhelming, Th1 response. In contrast, the immunopathology of NTM-PD may involve a more complex dysregulation, potentially including a Th1/Th2 imbalance, distinct cytokine profiles (e.g., high IL-13), and widespread T-cell exhaustion occurring in a different immunological context. However, these aspects have not been comprehensively mapped. The present study addresses this gap by leveraging high-dimensional mass cytometry and cytokine profiling, complemented by functional experiments, to comprehensively delineate the immune signature of peripheral blood in NTM-PD patients. Specifically, we aimed to identify alterations in immune subsets, cytokine profiles, and exhaustion markers, thereby uncovering potential immunopathological mechanisms underpinning disease persistence and progression. Clarifying these immune dysfunctions is critical, as it could reveal novel immunotherapeutic targets and biomarkers for predicting disease severity and treatment outcomes, addressing a crucial unmet clinical need in NTM-PD management.

Results

Immune profiling of PBMCs reveals distinct immune subset alterations in NTM-PD

We enrolled 28 newly diagnosed and treatment-naïve NTM-PD patients without immunosuppressive comorbidities (e.g., malignancies, type 2 diabetes, or autoimmune diseases) alongside 27 age-range- and sex-matched healthy controls (HCs) (Supplementary Data Cohort details, Supplementary Table 1). All participants were free of factors that might skew immune status. The control group was closely matched to the NTM-PD group in terms of gender, ethnicity, and smoking/alcohol history. Peripheral blood mononuclear cells (PBMCs) and plasma samples were collected from all subjects prior to treatment. High-dimensional CyTOF analysis (42-marker panel; Supplementary Table 2) and multiplex cytokine assays were performed to characterize disease-associated immune alterations (Fig. 1A, B; Supplementary Fig. 1A). Unsupervised clustering (FlowSOM) and t-SNE visualization revealed distinct immune-cell population patterns in NTM-PD patients versus controls (Fig. 1B; Supplementary Fig. 1A, B). Conventional flow cytometry on the same samples confirmed the CyTOF findings, showing concordant lymphocyte subset frequencies (Supplementary Fig. 2A–C). Both CyTOF and flow cytometry consistently showed an increased frequency of circulating monocytes and a decreased frequency of NKT-like cells in NTM-PD patients compared to healthy individuals (Fig. 1C). In particular, classical monocytes (CD14+ CD16−) were expanded in patient PBMCs, while CD3+ CD56+ NKT cells were reduced. These innate immune changes suggest a perturbed inflammatory baseline in NTM-PD.

Fig. 1. High-dimensional immune profiling reveals monocyte expansion and NKT-like reduction in NTM-PD.

A Study design: PBMC CyTOF and plasma Luminex in NTM-PD (n = 28) and HC (n = 27); patient etiologies MAC/MAB = 19/9; cytokine panel 48 analytes. B Merged and group-specific t-SNE visualizations of PBMCs, generated by FlowSOM clustering, show distinct immune landscapes in NTM-PD patients versus HCs, highlighting global shifts in cell population distribution. C Frequencies of classical monocytes (CD14+ CD16−) and CD3+ CD56+ NKT-like cells (median, IQR) demonstrate increased monocytes and decreased NKT-like cells in NTM-PD. Mann–Whitney U; multiple-comparison control as in the “Methods” section. (p < 0.05, *p < 0.01).

Identification of immune-cell subset changes in NTM-PD

Using the CyTOF dataset, we identified a total of 36 distinct immune-cell subsets in PBMCs, encompassing all major immune lineages (Fig. 2A, B; Supplementary Table 3). These included multiple subsets of monocytes, dendritic cells (DCs), natural killer (NK) cells, NKT cells, B cells, and 22 distinct CD3+ T-cell subsets. Heatmap and t-SNE analyses of these clusters (Fig. 2A, B, and Fig. 3B–E) allowed comparison between NTM-PD patients and controls, particularly highlighting T-cell compartment differences (Fig. 3B–E). A number of subsets showed significant frequency differences between NTM-PD and healthy individuals. In the innate compartment, classical monocytes (CD14+ CD16−, cluster M03) were markedly expanded in patients (Fig. 2C), whereas plasmacytoid DCs (cluster DC01) were significantly diminished (Fig. 2D). A specific NK cell subset (CD56loCD16loCD57-NK; cluster NK05) was also altered in patients (Fig. 2E). Consistent with initial observations, NKT-like cells (CD3+ CD56+, NKT) were significantly reduced in NTM-PD (Fig. 2F). Within the adaptive compartment, the analysis did not detect significant systemic alterations in the frequencies of naive or switched memory B cells, double-negative T (DNT) cell subsets, or Th17 cells (defined as CD3+ CD4+ CCR6+ CD161+ CXCR3−) (Fig. 3A, B). However, profound dysregulation was evident across other specific T-cell lineages. Notably, T helper 1 (Th1) cells (CXCR3+ CCR4−, cluster T03) were decreased in frequency, while T helper 2 (Th2) cells (CXCR3− CCR4+, clusters T04/T05) were modestly but significantly increased in NTM-PD patients (Fig. 3C). We also observed trends toward fewer regulatory T cells (Tregs, cluster T09) and more memory CD4+ T cells (cluster T07) in patients (Fig. 3C). Among CD8+ T cells, effector memory CD8+ T cells (clusters T12/T13) were significantly expanded, and both CXCR3+ and CXCR3− γδ T-cell subsets (clusters T18 and T22) were increased in NTM-PD compared to controls (Fig. 3D, E). These collective changes indicate a broad reshaping of the immune-cell landscape in NTM-PD, reflecting a shift toward a Th2-biased and chronically stimulated T-cell profile.

Fig. 2. Identification of altered immune-cell subsets associated with NTM-PD.

A Heatmap illustrating differential abundance of 36 identified immune subsets in PBMCs from NTM-PD patients and HCs. B t-SNE visualization highlighting significant differences in immune subset distributions between patient and control groups. Frequencies of significantly altered immune subsets, including classical monocytes (M03) (C), plasmacytoid dendritic cells (DC01) (D), CD56lo CD16lo CD57− NK cells (NK05) (E), and NKT cells (F). Data are expressed as the median (interquartile range). (p < 0.05, *p < 0.01 by Mann–Whitney U-test).

Fig. 3. Distinct alterations in T-cell subsets in NTM-PD.

t-SNE maps of CD4+ (A) and CD8+ (B) T-cell subsets, demonstrating the altered frequencies of T-cell populations in NTM-PD patients compared to HCs. C Quantitative comparisons showing decreased Th1 (T03) and increased Th2 subsets (T04, T05) in NTM-PD patients. Frequencies of effector memory CD8+ T cells (T12, T13) (D) and dysregulated γδ T-cell subsets (CXCR3+ γδT, T18; CXCR3− γδT, T22) (E). (p < 0.05, *p < 0.01, Mann–Whitney U-test).

Elevated expression of T-cell exhaustion markers in NTM-PD

Given the skewed T-cell subset distribution in NTM-PD, we next examined T-cell functional status by assessing key exhaustion markers. NTM-PD patients’ T cells exhibited significantly higher expression of inhibitory receptors, including PD-1, TIGIT, TIM-3, TCF1, and TOX, compared to controls (Fig. 4A; Supplementary Fig. 3). This upregulation was evident across multiple T-cell subsets. For example, even naive-like CD4+ T cells (cluster T01) in NTM-PD patients showed elevated PD-1 and TOX, and Th2 cells (cluster T05) had increased TIGIT and TIM-3 relative to controls. Effector memory CD8+ T cells (cluster T13) likewise displayed a notable increase in PD-1 and TOX, and unconventional γδ T cells (cluster T18) in patients exhibited an exhausted phenotype (Fig. 4A). A correlation analysis of immune-cell frequencies provided further insight (Fig. 4B). Plasmacytoid DC frequency was inversely correlated with Th2 cell frequency, while Th2 frequencies positively correlated with effector memory CD8+ T-cell frequencies (Fig. 4B). NKT cell frequency was positively correlated with Th1 cells and with γδ T cells. Overall, NTM-PD patients showed diminished anti-inflammatory Tregs and expanded pro-inflammatory/cytotoxic T-cell subsets that co-expressed high levels of exhaustion markers (e.g., PD-1, TOX), indicating an activated yet functionally impaired immune state.

Fig. 4. Enhanced expression of exhaustion markers on T-cell subsets in NTM-PD.

A Expression profiles of exhaustion markers (PD-1, TOX, TIGIT, TIM-3, TCF1) in selected T-cell subsets (naive CD4+ Th2, effector memory CD8+, γδT cells) from NTM-PD patients and HCs, revealing a pattern of expression in patients’ T cells that was largely opposite to that observed in healthy controls. B Correlation heatmap illustrating a curated selection of the most statistically significant and biologically relevant associations between immune subsets, including negative correlation between pDCs (DC01) and Th2 subsets (T04, T05), and positive correlation between Th2 subsets and effector memory CD8+ T cells (T12, T13) (p < 0.05, *p < 0.01).

Correlation of exhaustion markers in CD4+ and CD8+ T Cells

To further elucidate the network of T-cell exhaustion, we analyzed co-expression relationships among exhaustion markers in NTM-PD T cells. PD-1 and TOX were strongly positively correlated in both CD4+and CD8+populations (r = 0.52, p = 0.005 in each; Fig. 5A, B). Beyond this pair, multiple exhaustion markers (PD-1, TOX, TIGIT, TIM-3, TCF1, PD-L1) showed coordinated upregulation, forming a tightly interconnected inhibitory network (Fig. 5C). For instance, TOX expression correlated with TIGIT and TIM-3 as well as PD-1, and PD-1 levels correlated with TIM-3 and TIGIT in CD4+ cells and with PD-L1 in CD8+cells. TIGIT and TCF1 also correlated with TOX and PD-1. A heatmap of these correlations (Fig. 5C) highlighted PD-1 and TOX as central nodes. In summary, exhaustion markers were highly co-expressed: PD-1 levels strongly correlated with TOX in both CD4+and CD8+ T cells (Fig. 5A, B), and a broader analysis showed multiple inhibitory receptors rising in tandem (Fig. 5C). This indicates an orchestrated exhaustion program, with PD-1 and TOX appearing as central regulators of the exhausted T-cell phenotype.

Fig. 5. Correlation of exhaustion markers in CD4+ and CD8+ T cells from NTM-PD patients.

A, B Scatter plots indicating significant positive correlations between TOX and PD-1 expression in CD4+ (A) and CD8+ (B) T cells (Spearman correlation coefficients and p-values indicated). C Heatmap depicting interconnected correlation network among multiple exhaustion markers (TOX, PD-1, TIGIT, TIM-3, TCF1, PD-L1) within CD4+ and CD8+ T-cell compartments. Spearman correlation used for statistical assessment (p < 0.05, *p < 0.01).

Cytokine dysregulation in NTM-PD patients

NTM-PD was associated with broad cytokine dysregulation in peripheral blood. Multiplex plasma analysis showed that after correcting for multiple comparisons (FDR < 0.05), levels of 15 out of 48 tested analytes were significantly elevated in patients compared to controls (Fig. 6A). These included Th2-associated cytokines (e.g., IL-5, IL-13) and pro-inflammatory cytokines (e.g., IFN-γ, GM-CSF, IL-15), which were markedly higher in NTM-PD. Several chemokines and growth factors were also elevated, including CXCL9 (MIG), CXCL10 (IP-10), GRO-α, M-CSF, CTACK (CCL27), and VEGF. Notably, IL-13 was among the most dramatically elevated cytokines in patients. This pattern reflects a mixed inflammatory response: evidence of strong IFN-γ–driven activity (e.g., high CXCL9, CXCL10, IL-15) coexists with a pronounced type 2/tissue-remodeling signature (e.g., high IL-5, IL-13 and growth factors). To understand how these soluble mediators relate to immune-cell changes, we performed Spearman correlation analyses between cytokine levels and immune subset frequencies (Fig. 6B). Distinct relationships emerged. For example, IL-13, soluble IL-2Rα (sCD25), and CXCL10 levels positively correlated with classical monocyte frequency (cluster M03). Furthermore, effector memory CD8+ T-cell populations (clusters T12 and T13) also showed positive correlations with multiple cytokines, including VEGF and M-CSF, suggesting a link between systemic inflammation and the expansion of these cytotoxic cells. These findings underscore the central role of monocytes in orchestrating inflammation: high CXCL9/CXCL10 levels in patients, coupled with their correlation with monocyte counts, suggest monocyte-lineage cells are a major source of these chemokines in NTM-PD.

Fig. 6. Systemic cytokine alterations and their correlation with immune subsets in NTM-PD patients.

A Multiplex cytokine profiling revealed significantly elevated levels of cytokines and chemokines (e.g., IL-5, IL-6, IL-13, GM-CSF, IFN-γ, CXCL9/MIG, CXCL10/IP-10, and CXCL12) in NTM-PD patients compared to HCs. A color bar at the top distinguishes NTM-PD patients from healthy controls (HC). Full statistical analysis, including FDR-adjusted q-values for all 48 analytes, is provided in Supplementary Table S7. B Heatmap demonstrating significant Spearman correlations between plasma cytokines and immune-cell frequencies, including IL-13 and IP-10 positively correlating with classical monocytes, and VEGF, M-CSF, and IL-3 positively correlating with effector memory CD8+ T cells (p < 0.05, *p < 0.01).

Development of a predictive immune model for NTM-PD

The distinct immune signatures observed prompted us to investigate their utility as biomarkers. We employed recursive feature elimination combined with random forest classification to identify the most informative immune parameters distinguishing NTM-PD patients from controls (Fig. 7A). The top discriminatory features included cellular changes (elevated classical monocytes, cluster M03; elevated memory CD4+ T cells, cluster T07; and decreased NKT cells) alongside cytokine changes (e.g., high soluble IL-2Rα, CXCL9, IL-15, IL-6, GRO-α, CTACK). Other features, such as the frequency of M01 monocytes, also showed predictive potential but were not selected in the final, most parsimonious model. Using this refined feature set, we built a combined model integrating both immune cell and cytokine markers. The integrated model achieved an area under the ROC curve (AUC) of 0.922 (Fig. 7B), indicating excellent diagnostic accuracy. Notably, this combined model outperformed models based on cytokine features alone (AUC 0.869) or cell frequencies alone (AUC 0.725). Thus, a multiparametric immune profile can robustly classify NTM-PD status, and incorporating both cellular and soluble markers is crucial for optimal performance (Fig. 7A, B). Beyond diagnosis, our findings carried prognostic implications. Specifically, baseline plasma IL-13 levels were significantly higher in patients who failed to show radiographic improvement after six months of standard treatment compared to those who did improve (Fig. 7C). This observation aligns with reports that a Th2-biased inflammation corresponds to worse outcomes in severe infections³¹. It suggests that IL-13 could serve as a prognostic biomarker in NTM-PD: patients with highly Th2-skewed profiles may experience more progressive disease or respond less favorably to therapy.

Fig. 7. Development and validation of an immune-based predictive model for NTM-PD.

A Identification of the most informative immune features distinguishing NTM-PD from HCs using recursive feature elimination (RFE) and random forest analysis, including NKT cells, classical monocytes, effector memory CD8+ T cells, IL-2Rα, MIG (CXCL9), IL-15, GRO-α, IL-6, CTACK, TNF-β, IFN-γ, IL-8, β-NGF, and IL-3. B ROC curves comparing the diagnostic performance of combined immune markers (AUC = 0.922), cytokines alone (AUC = 0.869), and immune-cell subsets alone (AUC = 0.725). C Boxplot comparing baseline plasma IL-13 concentrations between NTM-PD patients with radiographic improvement (n = 18) and those without improvement (stable or worsened, n = 10) after 6 months of treatment. Statistical significance was determined by the Mann–Whitney U-test.

CXCL9/CXCL10 induction in macrophages

Our clinical data pointed to monocytes/macrophages as key contributors to the elevated CXCL9/CXCL10 levels in NTM-PD. To investigate a potential cellular source for the elevated CXCL9/CXCL10 levels observed in the NTM-PD cohort, murine BMDMs were infected with M. abscessus or MAC (ATCC 700898) at MOIs of 20 or 40. At 24 h post-infection, BMDMs infected with either NTM species secreted significantly elevated quantities of CXCL9 and CXCL10 proteins compared to uninfected controls. Although absolute secretion levels were lower than those induced by the LPS maximal activation benchmark, the NTM-induced chemokine release was highly robust (Fig. 8B). This chemokine induction was dose-dependent, with MOI 40 eliciting greater CXCL9/CXCL10 release than MOI 20. At the transcript level, infection with either M. abscessus or MAC, as well as LPS stimulation, incited a profound, dose-dependent upregulation of CXCL9 and CXCL10 mRNA (Fig. 8A), mirroring the protein secretion trends. These in vitro findings, benchmarked against classical pan-macrophage activators like LPS, definitively confirm that the profound induction of CXCL9/CXCL10 represents a highly conserved, potent innate macrophage response to diverse mycobacterial stimuli (both MAB and MAC). While this intense chemokine axis is not exclusively specific to NTM, it provides a crucial mechanistic link, confirming that activated monocyte/macrophage-lineage cells are a highly capable biological engine driving the extreme systemic chemokine gradients observed in the patients, continuously fueling the recruitment of T cells into the infected lungs.

Fig. 8. Macrophage-derived CXCL9 and CXCL10 innate responses to M. abscessus, MAC, and LPS stimulation.

A qPCR analysis of CXCL9 and CXCL10 mRNA expression in murine BMDMs infected with M. abscessus or MAC at MOIs of 20 and 40, or stimulated with LPS (100 ng/mL), demonstrating significant transcriptional induction correlating with chemokine secretion levels in a dose-dependent manner. B ELISA quantification of CXCL9 and CXCL10 secretion in supernatants of BMDMs following 24 h of infection with M. abscessus or MAC, or LPS stimulation, showing significant dose-dependent protein release compared to uninfected controls (p < 0.05, *p < 0.01, **p < 0.001, ***p < 0.0001).

Discussion

In this study, we delineated a comprehensive immune landscape for NTM-PD, revealing a convergence of immune hyperactivation and immune dysregulation that helps explain the pathogenesis of this disease. A hallmark finding is that NTM-PD is characterized by a misaligned immune response. Patients exhibit a paradoxical combination of innate immune hyperactivation with clear signs of IFN-γ signaling (e.g., expanded monocytes and high CXCL9/CXCL10 levels), alongside a skewed adaptive immunity defined by a Th2-biased environment (evidenced by elevated IL-13) and pervasive T-cell exhaustion^14,32,33. This profile diverges from the classic TB scenario—NTM-PD is not simply a milder form of TB, but an immunologically distinct entity in which Th1-associated inflammatory pathways appear active but functionally insufficient, coexisting with a non-protective Th2-dominated response.

A key feature of this immune dysregulation is the bias toward a Th2 phenotype, which appears to be functionally driven by IL-13. While the observed increase in circulating Th2 cell frequencies was modest, we found IL-13 levels to be profoundly elevated in NTM-PD plasma, correlating directly with monocyte counts (and inversely with Th1 frequency). This is consistent with previous reports of impaired IFN-γ (Th1) responses and compensatory Th2 shifts in NTM disease^34–36. This suggests IL-13 orchestrates a feed-forward loop: IL-13 (produced by Th2 cells and possibly innate lymphocytes) activates monocytes/macrophages, which then secrete factors that further reinforce Th2 polarization. IL-13 is known to drive alternative (M2) macrophage activation^14,37,38, leading to fibrosis and diminished microbicidal function. In mycobacterial infections, an IL-13–rich environment impairs macrophage ability to kill bacteria—indeed, TB models show that IL-13 overexpression worsens infection and pathology^39,40. Thus, excessive IL-13 in NTM-PD likely shifts immunity away from effective Th1-mediated clearance toward a Th2-dominated, non-protective state. This Th2 dominance may also contribute to the reduced regulatory T-cell (Treg) levels we observed, since IL-4/IL-13 can inhibit Treg development^41,42. Fewer Tregs would lift some brakes on inflammation—yet the unrestrained inflammation remains misdirected and ineffective at eliminating the pathogen. This suggests that in NTM-PD, the immune system is highly active but misdirected, emphasizing tissue remodeling over the necessary Th1-driven defense.

It is imperative to contextualize that the massive expansion of classical monocytes and the intense elevation of CXCL9/CXCL10 are not specific solely to NTM-PD; these features are well-documented hallmark innate responses to active pulmonary tuberculosis (PTB), driven by potent initial mycobacterial sensing^43,44. However, the critical systems-level distinction lies in the broader adaptive immunological network. In contained TB, these chemokines successfully recruit functional Th1 effectors. Conversely, the high-dimensional data in this NTM-PD cohort reveal a profound “immune misalignment.” The robust innate recruitment signals (CXCL9/10) paradoxically operate within a highly suppressive, Th2-skewed environment (driven by elevated IL-13) and are coupled with widespread T-cell exhaustion. The high CXCL9/CXCL10 likely draws T cells (e.g., Th1, CD8) into infected tissues, yet the persistent antigen load and Th2-skewed environment (high IL-13, IL-5) drive those recruited T cells into exhaustion rather than effective immunity. Consistent with this, we observed high IFN-γ–induced chemokines alongside low circulating Th1 cells in patients, and widespread upregulation of exhaustion markers (PD-1, TOX, TIGIT, TIM-3) on their T cells. Notably, TOX—a master transcriptional regulator governing irreversible T-cell exhaustion^18–20—was highly expressed across T-cell subsets. The unexpected and robust elevation of PD-1 and TOX on naïve-phenotype T cells (cluster T01) warrants specific mechanistic attention. While the upregulation of TOX and PD-1 is classically driven by chronic T-cell receptor (TCR) engagement, recent immunological advances demonstrate that inflammatory cytokines (e.g., IL-6, IL-12, IL-15, and IL-18) are sufficient to induce TOX and PD-1 expression in a TCR-independent manner. Given the severely dysregulated cytokine profile observed in the NTM-PD cohort (Figs. 6 and 7), our analysis suggests that naïve T cells are subjected to profound cytokine-driven bystander activation, forcing premature epigenetic and transcriptional imprinting of exhaustion networks prior to specific antigen encounter. NKT cells, which normally provide early IFN-γ against mycobacteria but can also produce Th2 cytokines^45–47, were markedly reduced in patients. Their absence likely hinders early bacterial control, and any IFN-γ/IL-13 they produce could both induce chemokines and contribute to Th2 skewing. Together, these findings support a model of an “immune trap” or “ineffective inflammation”: the IFN-γ–CXCL9/CXCL10 axis initiates immune recruitment, but the hostile, IL-13-rich milieu renders the responding T cells dysfunctional. This explains why NTM-PD patients have robust immune activation yet fail to clear the infection—the immune system is essentially wearing itself out (via exhaustion) in the process of responding to the pathogen.

While both NTM-PD and TB can exhibit T-cell exhaustion in their chronic forms, their immune regulation differs significantly¹⁷. TB typically elicits a strong Th1 response and engages regulatory mechanisms (e.g., expanded Tregs and PD-L1 upregulation) to prevent excessive inflammation. In contrast, our data show a trend towards reduced Tregs in NTM-PD, indicating an inflammatory milieu that is less restrained but ultimately rendered ineffective via exhaustion within a Th2-biased setting. In NTM-PD, the immune system mounts a vigorous response (as evidenced by high circulating cytokines), yet this response is misaligned and unsustainable. Thus, NTM bacteria evade clearance not by escaping immune detection, but by driving the host immune response into a state of dysfunctional activation—where immune cells are present and activated but too exhausted to eliminate the infection.

These immune alterations have potential clinical utility. The combination of high IL-13 and soluble IL-2Rα levels, elevated CXCL9/CXCL10, expanded monocytes, and diminished NKT cells could serve as an immunological “fingerprint” for NTM-PD. In our cohort, a model using such markers distinguished NTM-PD patients from controls with high accuracy and even predicted treatment outcomes—patients with the most Th2-skewed profiles (e.g., high IL-13) had poorer radiographic improvement. Therefore, immune biomarkers like IL-13 may identify patients at risk of severe or unresponsive disease, who might benefit from more aggressive or adjunctive therapy.

Our results also suggest new therapeutic approaches. Targeting the IL-13/Th2 pathway—using IL-13 neutralizing antibodies or IL-4Rα blockers (already available for allergic diseases)—could rebalance the immune response in NTM-PD by enhancing Th1 activity and improving macrophage bactericidal function. Conversely, relieving T-cell exhaustion via immune checkpoint inhibitors (e.g., anti-PD-1 or anti-TIGIT) might reinvigorate the immune response to NTM. Preclinical studies in TB models show that PD-1 blockade can boost T-cell function and reduce mycobacterial burden, although unchecked T-cell activation can cause pathology (PD-1 knockout mice succumb to TB due to hyper-inflammation). Any immunotherapy in NTM-PD would therefore need to be carefully controlled—potentially localized or timed with antibiotic treatment—to avoid excessive inflammation. Short of therapy, monitoring exhaustion markers in patients could be clinically useful: a decline in PD-1highT cells during therapy may indicate immune recovery and forthcoming bacterial clearance, whereas persistently high levels might warn of ongoing infection and treatment failure¹⁷. Such immune-monitoring strategies could complement standard microbiological assessments in NTM-PD management.

In summary, this study provides a comprehensive peripheral immune landscape of NTM-PD pathogenesis. The data delineate a vigorously activated yet potentially misdirected systemic immune network. This specific immunological signature provides a mechanistic hypothesis regarding impaired bacterial clearance despite extensive systemic immune activation. Importantly, our findings highlight measurable biomarkers (such as elevated IL-13, CXCL10, and PD-1highT cells) that could aid in diagnosis or prognosis and suggest novel avenues for therapy (e.g., targeting the IL-13/IL-4 axis or T-cell exhaustion pathways). Future studies should verify these immune mechanisms in lung tissue and test whether correcting the Th1/Th2 imbalance or reversing T-cell exhaustion can improve outcomes. Overall, deciphering this immune dysregulation moves us closer to tailored, host-directed strategies to enhance antimycobacterial treatment. Restoring Th1 immunity and relieving T-cell exhaustion may be key to enabling the host to control NTM-PD, which is increasingly important as NTM infections rise globally.

However, these findings must be interpreted in the context of several limitations. First, the relatively small sample size (n = 28 NTM-PD patients) and the restriction of analysis to peripheral blood mononuclear cells (PBMCs) preclude a definitive understanding of localized host-pathogen dynamics within the pulmonary granulomatous tissue. Second, the observational nature of the CyTOF dataset lacks functional validation via antigen-specific effector T-cell assays. Third, the mechanistic in vitro experiments relied on a murine BMDM model; future investigations must employ homologous human primary innate immune cells to account for species-specific differences. Fourth, the reliance on a healthy control group, without a parallel cohort of patients with active pulmonary tuberculosis (PTB), limits the capacity to definitively separate NTM-specific immunopathology from general mycobacterial inflammatory responses. Finally, while the predictive model demonstrates high accuracy within this dataset, it requires prospective validation in independent multi-center cohorts.

Methods

Ethical statement

This study was approved by the Zhongshan Hospital Ethics Committee, Fudan University (B2024-021) in accordance with the Declaration of Helsinki, and written informed consent was obtained from all participants.

Study participants

This study enrolled 28 treatment-naïve NTM-PD patients and 27 age- and sex-matched healthy controls. All patients lacked immunosuppressive comorbidities (e.g., malignancy, diabetes, autoimmune disease) and met rigorous diagnostic criteria for NTM-PD in strict accordance with the 2020 ATS/ERS/ESCMID/IDSA Clinical Practice Guidelines. Microbiological confirmation necessitated the isolation of the identical NTM species from at least two independent expectorated sputum samples collected over an interval of a week or more. Samples were subjected to in vitro enrichment via the BACTEC MGIT 960 liquid culture system. Definitive species classification was performed using Matrix-Assisted Laser Desorption/Ionization Time-of-Flight Mass Spectrometry (MALDI-TOF MS). Critically, co-infection with the Mycobacterium tuberculosis complex was definitively ruled out in all subjects through concurrent negative GeneXpert MTB/RIF testing and negative specific MTB cultures.

Sample collection and preparation

Approximately 10 mL of peripheral blood was collected per participant. PBMCs were isolated within 2 h by Ficoll-Paque density gradient centrifugation, washed, and cryopreserved in FBS with 10% DMSO in liquid nitrogen for later analysis.

Immune profiling using mass cytometry (CyTOF)

Mass cytometry (CyTOF) was performed on PBMCs using a 42-marker metal-conjugated antibody panel covering major immune lineages, activation markers, exhaustion markers, cytokine receptors, and transcription factors. PBMCs were thawed, counted, and stained with surface antibody cocktails (30 min, 4 °C). After washing, cells were fixed (1% formaldehyde), permeabilized with methanol, and stained with intracellular antibodies. Ir-191/193 DNA intercalator was added for live/dead discrimination, and data were acquired on a Helios CyTOF (Fluidigm).

Flow cytometry analysis

Conventional flow cytometry was also performed on thawed PBMCs. Cells were stained with fluorochrome-conjugated antibodies (Supplementary Table 4) and a viability dye for 25 min at 4 °C, washed, and analyzed on a BD FACSymphony A3 flow cytometer (BD Biosciences). Data were processed using FlowJo v10.

Cytokine profiling

Plasma cytokine concentrations were measured using a 48-plex Luminex bead-based assay (R&D Systems) according to the manufacturer’s instructions. Data were acquired on a Bio-Plex 200 system and analyzed with Bio-Plex Manager software.

Radiographic assessment and treatment

Following baseline immunophenotyping, patients received targeted antimicrobial therapy guided by CLSI susceptibility standards. MAC patients predominantly received a standard multidrug regimen (macrolide, ethambutol, and rifampin), whereas MAB patients required intensive tailored regimens (typically a macrolide combined with intravenous amikacin and cefoxitin) due to their complex resistance profiles. Baseline and 6-month follow-up chest CTs were obtained for all 28 patients and independently reviewed by two board-certified thoracic radiologists blinded to clinical metadata. Changes were categorized as improved, stable, or worsened based on the overall burden of bronchiectasis-related abnormalities (e.g., tree-in-bud nodules, consolidations, cavities) and their extent/intensity versus baseline; discrepancies were resolved by consensus.

BMDM culture and infection protocol

All animal procedures were conducted in strict accordance with the ARRIVE guidelines and the institutional regulations for the care and use of laboratory animals. The experimental protocols were reviewed and approved by the Institutional Animal Ethics Committee of Ruijin Hospital, Shanghai Jiao Tong University School of Medicine (Approval No. JUMC2023 046 B). Female C57BL/6J mice (8 weeks old) were housed under specific pathogen-free (SPF) conditions, maintained on a standard 12-h light/dark cycle, and provided ad libitum access to standard chow and autoclaved water. As the animals were solely utilized as tissue donors for primary cell isolation rather than subjects for in vivo therapeutic interventions, in vivo group randomization and blinding were not applicable.

To minimize animal distress, mice were humanely euthanized via CO2 asphyxiation followed by cervical dislocation. Bone marrow was immediately harvested from the femurs and tibias and cultured in DMEM supplemented with 10% FBS and 50 ng/mL M-CSF for 4 days to generate BMDMs. Adherent macrophages were harvested and seeded at a density of approximately 1 × 10⁶ cells/well in 12-well plates. To provide a genetically homogeneous and pathogen-naïve phagocytic baseline, minimizing inter-donor variability, these murine BMDMs were utilized for downstream assays. After 12 h of resting, BMDMs were infected with M. abscessus (MAB, ATCC 19977) or Mycobacterium avium complex (MAC, ATCC 700898) at a multiplicity of infection (MOI) of 20 or 40. Additionally, parallel wells were stimulated with Lipopolysaccharide (LPS, 100 ng/mL) as a potent pan-macrophage activating positive control. All conditions were incubated for 24 h alongside uninfected media controls. Following the 24-h incubation, culture supernatants and cell pellets were collected for downstream analysis.

CXCL9/CXCL10 measurement

CXCL9 and CXCL10 protein levels in BMDM supernatants were measured by ELISA using mouse-specific kits. In parallel, total RNA from BMDM pellets was isolated (e.g., TRIzol), and cDNA was synthesized from 1 μg RNA. CXCL9 and CXCL10 mRNA expression was quantified by qPCR (primers in Supplementary Table 5) normalized to the ACTB housekeeping gene using the 2 − ΔΔCt method.

Data and statistical analysis

Unsupervised clustering was performed with FlowSOM, and dimensionality reduction with t-SNE in R. Unless otherwise specified, two-tailed tests were used. Between-group comparisons were analyzed primarily with the unpaired Mann–Whitney U-test; where data approximated normality and homoscedasticity on inspection, unpaired t-tests (two groups) or one-way ANOVA (≥3 groups) were used. When ≥3 groups did not meet parametric assumptions, the Kruskal–Wallis test with Dunn’s post-hoc comparisons was applied. Correlations were assessed using two-tailed Spearman’s rank.

For families of simultaneous tests (e.g., across multiple immune-cell subsets, multiplex cytokines, or entries within correlation matrices), multiple comparisons were controlled using the Benjamini–Hochberg false discovery rate (FDR); FDR-adjusted q-values are reported in the text/legends and Supplementary Tables where applicable. To preserve the original figures, stars shown in figures denote nominal p-values; the corresponding FDR results are stated in the legends or the “Methods” section. A significance threshold of α = 0.05 was used unless otherwise indicated.

For classification analyses, we used a random forest model coupled with recursive feature elimination (RFE) and 5-fold cross-validation. Discrimination was summarized by the area under the ROC curve (AUC); when AUCs were compared, DeLong’s method was used. Data are presented as median (interquartile range) unless noted; exact sample sizes (n) denote independent biological subjects and are specified in the figure legends. All analyses were conducted in R and standard statistical software.

Author contributions

Tingting Fang: Writing – original draft, Investigation. Feifei Yuan: Writing – original draft, Investigation. Yu Chen: Writing – original draft, Investigation. Shuang Wan: Methodology, Data curation. Na Li: Methodology, Data curation. Yuyan Ma: Methodology, Data curation. Qingqing Wang: Methodology, Validation. Wenting Jin: Methodology. Qing Miao: Methodology. Lijuan Hu: Validation. Jue Pan: Validation. Donghe Li: Writing – review & editing, Project administration, Conceptualization. Bijie Hu: Writing – review & editing, Project administration, Funding acquisition, Conceptualization. All the authors reviewed and approved the final version of the manuscript.

Data availability

The datasets generated and/or analyzed during the current study are not publicly available due to their inclusion in an ongoing study and planned use in future analyses, but are available from the corresponding author on reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41540-026-00784-2.